The distributions of certain record statistics from a random number of observations (Q811048)

Let \(Y_ 0,Y_ 1,...,Y_ N\) be a sequence of i.i.d. random variables, N a non-negative integer-valued r.v., independent of \(Y'_ i s\), \(P(N=k)=p_ k\) with pgf \(H(x)=\sum^{\infty}_{k=0}x^ kp_ k\), \(x_ 0,x_ 1,..\). a sequence of record values, the successive maxima. \(D_ n\) denotes the event that at least n records occur, and \(E_ n\) the event that exactly n records occur, \(n\geq 0\), \(s_ j=F(x_ j).\) This paper deals with the above concepts in four sections. Section 1 sets the stage with preliminaries and definitions. In Section 2, the authors give five lemmas out of which two are proved, all dealing with \({\mathfrak s}=(s_ 0,s_ 1,...,s_{n-1})\). Section 3 gives the representation for g(\({\mathfrak s})\) through 3 theorems and 2 lemmas, where the joint distribution of the first n record values, if at least N records occur, is \[ P_{n,N}=P[D_ n\cap X_ 0\in dx_ 0...X_ n\in dx_ n| \quad N=k]=\prod^{n}_{j=0}ds_ jg({\mathfrak s}). \] All three theorems and one of the lemmas are proved. The last Section 4 consists of distribution theory with expressions for \(P_{n,N}\). There are five theorems in this section of which three are proved in detail. One of these gives the representation for \(P_{n,N}\) for the strong records in the i.i.d. model or in the Pfeifer model [\textit{D. Pfeifer}, J. Appl. Probab. 19, 127-135 (1982; Zbl 0484.60013); Corrections in ibid., 906-907 (1982)]. Another theorem deals with the event \(E_ n\) that exactly n records occur, by giving expressions for \(P[E_ n\cap X_ 0\in dx_ 0,...,X_ n\in dx_ n]\). The last result in this section is that \(P(E_ n)\) has a shifted truncated Poisson distribution under certain conditions.